几种图像缩放算法的研究

Research on image scaling algorithm

时域内常用的几种图像缩放算法有:最近邻插值、线性插值、二次插值、三次插值、拉格朗日插值、高斯插值等,对这些算法的性能进行分析比较,综合通带、阻带及截止频率,最近邻插值和线性插值应该避免,高斯基函数(N较大者)具有较好性能;并且在频域内研究了二维可分离插值滤波器和不可分离插值滤波器,这两种方法以对通带和阻带的要求作为优化目标,以滤波器的结构为约束条件,将滤波器的设计转化为一个约束优化问题进行解决;实验结果表明二维不可分离插值滤波器的方法图像缩放后的效果最好。

There are several kinds of image scaling methods in time domain： nearest neighbor interpolation,linear interpolation,quadratic interpolation,cubic interpolation,Lagrange interpolation,Gaussian interpolation,etc.The performance of these methods are compared according to passband,stopband and cutoff frequency.The result shows that the nearest interpolation and the linear interpolation should be avoided,and Gaussian basis function whose N value is larger has better perfor-mance.The 2-D nonseparable interpolation filter and 2-D separable in-terpolation filter are studied.With the requirements of passband and stopband as optimizing target,and the struc-ture of filter as constraint condition,the filter design for them is converted into an optimization problem.The ex-periment result shows that the 2-D nonseparable interpolation filter has better performance.